Analogue clipping from digital sources


Given the high output (typically >2v in RCA, >4v in balanced mode) there is in my experience a significant risk of overloading both the analogue stage of the DAC and any pre- or power amp downstream. Fighting that with low volume settings on the attenuator only further aggravates the issue. In my case I have to run the InnuOS Zenith’s Mk3 output at 85% to deal with audible overloading of the DAC/ amp with audible distortion. Anyone with similar experience?

antigrunge2

Hey,

Have you actually heard this happening, or are you just theorizing?

There is risk of digital clipping, but it’s not from the output voltage. It’s from upsampling. There’s a good paper on it from Benchmark Media somewhere about what happens.

Imagine a 44.1kHz/16 bit recording. Let’s say it’s a simple sine wave and the original samples happen to reach maximum. At the top, you could have two samples at peak output in a recording. Not an uncommon thing as many mastering engineers push the max sample to the peak digital output to ensure the recording has plenty of dynamic range.  Depending on the precise timing, you could have two adjacent peak signals or one very close and the other at peak.

Anyway, if you convert from 44.1/16 to 88.1/16 with linear interpolation there will be no issue, but most upsampling (thanks to cheap compute power) use something like a French curve to interpolate.  Now our new  samples can exceed maximum digital output. That is true digital clipping. The solution is to slightly reduce the signal when upsampling. Not a horrible thing when you also increase the bit depth before doing so.

I forgot one important point. Most preamps can take significant overload on the inputs.

Amplifier output is tied to input and usually 20x the input. So long as you don’t exceed the maximum output voltage they will take more than 2V in.  Stereophile may state the "maximum input voltage before clipping" or something like that.

Thank you for this: I am using upsampling, clearly heard clipping like noise and the cure is to lower the output from the server. Learned something new!

Technically, the best solution is to lower the output after bit depth conversion (16 to 24 or 32 bits) but before upsampling.

Roon calls this setting Headroom Management.

The end result of this 2-step dance is you can reduce the maximum output without reducing resolution while minimizing how much about the original recording you must know to avoid clipping.

I am on InnuOS Sense feeding into the Antelope Zodiac Platinum DAC and believe I can only use the Sense imbedded attenuator. Any other suggestion? Many thanks for your help, much appreciated!

BTW: this subject should find a wider audience.